resharp-wasm

A regex library for JavaScript, built on resharp. It supports intersection, complement, and lookarounds, and always runs in time proportional to the input length (no catastrophic backtracking).

For pattern syntax and supported features see the resharp README.

Pre-1.0: the API may change between minor versions.

Install

npm install @ieviev/resharp

Works in Node and with bundlers (Vite, webpack, rollup, esbuild).

Usage

import { Regex } from "@ieviev/resharp";

const r = new Regex("\\d+");
r.isMatch("abc 42");                    // true
Array.from(r.findAll("a12 b34 c5"));    // [1, 3, 5, 7, 9, 10]   string positions
Array.from(r.findAllBytes("αX"));       // [2, 3]                byte positions

Each match is reported as a pair of numbers: where it starts and where it ends. findAll counts in string positions (the same units JavaScript strings use). findAllBytes counts in bytes of the UTF-8 encoding, which is useful when working with Uint8Array data.

Inputs can be a string or a Uint8Array. By default \w, \d, \s, and . follow JavaScript's built-in regex (without the u flag).

Options

The constructor accepts an optional RegexOptions object.

new Regex("hello", { caseInsensitive: true });
new Regex("a.b",   { dotMatchesNewLine: true });
new Regex(pat, {
    mode: "js",              // "js" | "resharp" | "full"
    caseInsensitive: false,
    dotMatchesNewLine: false,
    ignoreWhitespace: false,
    hardened: false,         // prevents adversarial blowup
    dfaThreshold: 0,         // states to eagerly precompile
    maxDfaCapacity: 65535,   // cached DFA states
    lookaheadContextMax: 800,
});

Unknown keys and wrong-typed values are rejected by the constructor.

Cleanup

Each Regex holds memory outside of JavaScript's garbage collector. Release it with r.free(), or let using do it for you:

{
    using r = new Regex("\\d+");
    r.isMatch("42");
}

API

class Regex {
    constructor(pattern: string, options?: RegexOptions);
    isMatch(input: string | Uint8Array): boolean;
    findAll(input: string | Uint8Array): Uint32Array;                   // string positions
    findAllBytes(input: string | Uint8Array): Uint32Array;              // byte positions
    findAnchored(input: string | Uint8Array): Uint32Array | undefined;  // string positions
    free(): void;
    [Symbol.dispose](): void;
}

Limitations

No capture groups. No replace or split. Results give you only the positions of each match. The constructor throws on invalid patterns; matching can throw if the engine hits an internal size or number limit.

The compiled bundle is around 900 KB, mostly Unicode tables. A smaller ASCII-only build is planned.

Build

./build.sh    # build via nix shell
nix build     # pinned toolchain, ./result is the npm layout
nix develop   # devshell

License

MIT